Standing-up and sitting-down action detection method based on face position information

Abstract

A method for detecting standing and sitting actions based on face position information relates to the technical field of image processing, and comprises the following steps: constructing a feature space for outlier detection; constructing pseudo depth information; constructing a pseudo three-dimensional coordinate as a face feature; performing a boundary learning task by using a linear regression algorithm to learn an optimal plane; for a multi-person standing occasion, matching tracking is carried out on detection objects of front and back frames through an optimal transmission theory, and the function that the objects are stably tracked while multiple persons stand is achieved. According to the method, key spatial features are enhanced, the recognition effect of action detection is improved, the accuracy and universality of computer vision recognition are improved, and the real-time efficiency of detection is improved.

Description

technical field [0001] The invention relates to the technical field of image processing, and in particular, to a method for detecting standing up and sitting down actions based on face position information. Background technique [0002] At present, most standing detection algorithms are implemented based on deep learning gesture recognition or target tracking algorithms. Among them, the algorithm of gesture recognition based on deep learning usually requires high calculation amount of GPU and CPU, and the number of parameters of the model is also large. Therefore, deploying it in classroom teaching or lecture halls and other scenarios to realize real-time standing detection of all objects in the crowd has high requirements on the computing resources of the device. The standing detection based on the tracking algorithm needs continuous video stream as input, the front and back frames have a great influence, cannot achieve independent single-frame detection, and has poor abil...

AI Technical Summary

Problems solved by technology

The standing detection based on the tracking algorithm requires continuous video stream as input, and the front and rear frames have a great influence, so independent single-frame detection cannot be achieved, and the ability to recognize high-speed moving targets is poor, and the input based on video stream will reduce the size of the input data. It is larger than a single frame input, and the related pre-processing operations have a large amount of calculation

Images